Design of an In-Process Quality Monitoring Strategy for FDM-Type 3D Printer Using Deep Learning
, Danielle Jaye S. Agron 2, Gabriel Chukwunonso Amaizu 3, Dong-Seong Kim 2 and Jae-Min Lee 2,*
Round 1
Reviewer 1 Report
The material is hot, and its viscoelastic behaviour causes the "cut" of the corner: in the presence of a corner, the material does not precisely follow the path of the nozzle but tends to cut the corner introducing a shape error. This aspect has not been investigated. Explain the motivation
when the nozzle changes direction, there is a variation in the speed components; the viscoelastic behavior of the material causes a non-linear response and a delay time before reaching the stationary condition with the consequent local variation of the dimensions of the deposited material. Does the proposed solution eliminate this source of residual stress?
It is the opinion of the reviewer that the paper investigates only the macro and catastrophic defect and does not contribute to increment the knowledge of the process. the author must justify their choice to evaluate only some aspects of the complex problem
Author Response
To Associate Editor:
Thanks for your kind comments. Your letter indicated that our manuscript
(applsci-1849536) should be revised in order to answer certain issues that have
been addressed by the reviewers. The paper was revised and improved according to
the reviewers' comments.
The followings are the revised contents and answers (italics) to each reviewers'
suggestions.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The manuscript presents the development of a deep learning algorithm built towards real-time monitoring of the FDM additive manufacturing process. The authors claim that the developed deep learning algorithm, by continuously assessing printing conditions (such as temperature of the extruder motor, vibration and acoustic emission) is able to predict malfunctioning with increased performance when compared to other prominent algorithms.
Even though the study is interesting and the results worth of being published, the reporting can be confusing and major amendments are necessary to improve the explanation of the algorithm validation and overall quality of this study presentation.
Major revisions:
1.
The way the algorithm may predict malfunctioning is not clear, since there is no overview/description of the actual printing conditions nor comprehensive definition of the criteria for functioning/malfunctioning of the additive manufacturing process. An overview of the printing experiment should be included for better understanding of the considered dataset.
2.
Temperature, printing activity, vibration and acoustic emission are listed as parameters. However, their signal within the FDM additive manufacturing process is not shown. The authors should better describe the correlation between the deep learning algorithm and the measured sensors output. Results should include the amplitudes of the registered acoustic emissions, temperatures, and vibration.
3.
Even though the authors claim the development of a monitoring system capable of predicting anomalous behaviors, that is not clearly shown in the results. The authors are encouraged to further illustrate such events with practical examples, using the developed deep learning algorithm.
4.
Lines 67-70. This description of each section is unadjusted. The section numbers do not correspond to the actual section of the paper. The authors should be more specific and inclusive. For instance: when referring to “data gathering” authors are encouraged to specify what data.
5.
Lines 12-70: In this introductory text, both literature review and some explanation of current work is simultaneously conducted. Their distinction is sometimes not clear. Current research work should be moved to last paragraph (lines 67-70) where a description of the research work is conducted.
7.
Figure 1 shows the experimental setup of this research work. The authors are encouraged to elaborate on this topic and present a better description of the sensor systems. Figure 1 should be divided in subfigures (a, b, c,…) and moved to the experimental setup.
8.
The authors conclusions are, in reality, future work which gives the idea of unfinished research. The authors are encouraged to include at least some kind of preliminary testing of the 3D printing operational monitoring/tracking and briefly conclude on the developed methodology.
Minor revisions:
-line 1: please do not refer to additive manufacturing as a scheme. Some alternatives could be “processes”; “technologies”, …;
The same applies for the title of the paper. Suggestions for replacement: “methodology”; “strategy”, …
-lines 3-5: The sentence seems unnecessarily long. Please revise.
Suggestion: The development of a real-time process monitoring system with the ability to properly forecast anomalous behaviors within fused deposition modeling (FDM) additive manufacturing is proposed as a solution to the particular problem of nozzle clogging.
-line 6: “and processed” to “and its processing”
-line 6: “accumulate” --> “collect”
-line 7: “various features”. Please be more specific and concise (which features?).
-line 12: The creation of a subsection called “1. Introduction” is suggested.
-lines 12-23: Quite repetitive ideas. Please simplify.
“users have to wait a couple of hours”
“using 3D printers in a couple of hours”
“it is time-consuming and”
-line 34: “These” --> “This”
-line 83: “In this section, we will first discuss”. Please avoid using the first-person pronouns in scientific writing.
-line 91: Dx notation is not shown in equation 1.
-line 135: data collection frequency of 2047 Hz
-line 124: “Experiment” --> “Experimental”
-line 141-142: Table 2 seems unnecessarily long and most related with literature review. Is this an outcome of the following paper? How come it does note have at least one reference?
The whole document should be reviewed for English language.
Author Response
To Associate Editor:
Thanks for your kind comments. Your letter indicated that our manuscript
(applsci-1849536) should be revised in order to answer certain issues that have
been addressed by the reviewers. The paper was revised and improved according to
the reviewers' comments.
The followings are the revised contents and answers (italics) to each reviewers'
suggestions.
Author Response File: Author Response.pdf
Reviewer 3 Report
The manuscript attempts to present a case about the design of a in-process quality monitoring scheme for FDM-type 3D Printers by using Deep Learning algorithm. The paper is well-written in terms of use of English and has the style and language demanded for a potential publication. The sensors proposed to be used, have indeed some interest, especially the acoustic one. However, the methodology has some mistakes, especially in the understanding of the problems caused in the extruder by the selected temperature. The conduction of the aforementioned corrections is essential for the paper to have scientific soundness. The list of references should definitely be expanded, since they are very few in quantity.
My points are analytically listed below
Points for consideration:
• Point 1: In lines 12-13, please rephrase the sentence and insert the words “and resins” after the word “metal”.
• Point 2: In line 37, please rephrase the sentence and insert the words “and end-use products” after the word “prototypes”.
• Point 3: In lines 23, please also include the following reference about 3D Printing failure rates:
· • 10.1109/ICIEA.2019.8834376
• Point 4: In line 33, please also include the following relevant papers about 3D Printing process monitoring via using sensors:
· • 10.1016/j.matdes.2013.02.067
h•https://www.diva-portal.org/smash/get/diva2:660817/FULLTEXT09.pdf#page=163
· Point 5: In lines 129-130, authors mention that “If the temperature reaches below 220°C, then there is a high possibility that the nozzle starts to clog on the nozzle tip”. This is an absolutely wrong statement, at least regarding the use of PLA material. There are many known PLA filaments where manufacturer manuals instruct users to print in temperature starting even at 1800C. Please rephrase and reconsider your methodology.
· Point 6: The selected temperature for the extruder can have various implications. A low temperature can cause clogging due to material partially not melting, while high temperatures can cause the so called “heat crip” phenomenon, causing again clogging. This also should be taken into account.
· Point 7: Have authors considered the case where an extruder fan might fail, cause a heat crip phenomenon and clog the extruder? This is something common that could also be detected by the acoustic sensor.
• General point: Since authors propose the use of Deep Learning algorithm, have they considered mentioning the concept of Digital Twins? I believe that a paragraph about it would be worth to be written.
The reviewer.
Author Response
To Associate Editor:
Thanks for your kind comments. Your letter indicated that our manuscript
(applsci-1849536) should be revised in order to answer certain issues that have
been addressed by the reviewers. The paper was revised and improved according to
the reviewers' comments.
The followings are the revised contents and answers (italics) to each reviewers'
suggestions.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Author Response
We thank you for helping us improve this paper. We highly appreciate it.
Reviewer 2 Report
I suggest the word "algorithm" is either removed from the title. Alternatively, including an definite article could also be a solution: "using a deep learning algorithm"
Please refrain from using first person pronouns (such as "we") in scientific writing.
Author Response
Thanks for your kind comments. Your letter indicated that our manuscript
(applsci-1849536) should be revised in order to answer certain issues that have been addressed by the reviewers. The paper was revised and improved according to the reviewers' comments.
The followings are the revised contents and answers (italics) to each reviewers' suggestions.
Author Response File: Author Response.pdf
Reviewer 3 Report
I would like to thank the authors for their revisions. They have made a fine job and I believe that the manuscript is almost ready for publication.
One last point:
Regarding my last point about the Digital Twins, please add the following two relevant references about Digital Twins in 3D Printing:
10.1016/j.scriptamat.2016.12.005
10.3390/asi5010007
After these two add ons, I believe that the manuscript will be ready for publication.
Author Response
Thanks for your kind comments. Your letter indicated that our manuscript (applsci-1849536) should be revised in order to answer certain issues that have been addressed by the reviewers. The paper was revised and improved according to the reviewers' comments.
The followings are the revised contents and answers (italics) to each reviewers' suggestions.
Author Response File: Author Response.pdf
